Duolingo (DUOL): the learning app that turned free users into a cash machine

Duolingo passes both hard filters: ROE of 38.1% (FY2025) and a 3-year FCF CAGR of ~97% (FY2022–FY2025, from $48M to $370M). The stock has fallen 79% from its 52-week high as the market frets over slowing bookings guidance and a deliberate engagement-first strategy shift. That selloff has compressed the multiple to levels not seen since 2022 — 12.5x TTM P/FCF and 22.7x EV/EBITDA — while the underlying cash engine keeps compounding at rates that are hard to find in large-cap software.

Both filters cleared from verified annual financials. 1 2

The ROE number deserves a note: it inflated sharply in FY2025 because a deferred tax asset recognition added $231M to net income (negative tax provision). Strip that out and operating-level earnings were roughly $183M — still producing an adjusted ROE north of 20% on a growing equity base of $1.35B. The cash flow statement is unaffected by the tax item, so the FCF trajectory ($370M, 35.6% margin) is the cleaner picture of economic output.

Duolingo runs 250 language courses through a single mobile app that has 56.5 million daily active users and 12.5 million paid subscribers as of Q1 2026. 3 Revenue is split across three tiers: a free ad-supported tier, Duolingo Plus (~$7/month), and Duolingo Max (~$14/month, with AI-driven conversation practice and unlimited hearts). The company also sells the Duolingo English Test (DET), a certified language proficiency exam accepted by over 5,000 institutions globally as an alternative to TOEFL and IELTS.

Revenue compounded at a 41% rate from FY2022 to FY2025, reaching $1.04B. Q1 2026 revenue was $292M, up 27% year-on-year, with DAUs growing 21% — the first quarter where DAU growth outpaced revenue growth by design. Full-year 2026 guidance is for $1.205B in revenue and ~$310M in adjusted EBITDA, implying a 25.7% EBITDA margin. 4

Three things protect this business that are easy to see but hard to replicate:

Habit and streak mechanics. Duolingo's gamification layer — streaks, XP, leaderboards — is so psychologically sticky that 500+ streak days are common among paid users. Switching to a competing app means starting over socially and psychologically. This isn't just feature design; the streak data is what trained Duolingo's AI models, giving it a dataset advantage that any new entrant would need years to rebuild.

Content depth at scale. 250 language courses covering grammar systems, scripts, and tones across dozens of language families took over a decade to build. The curriculum is layered to A1 through C1 levels. In April 2026, Duolingo opened advanced content — previously a paid-tier feature — to free users across nine languages, an offensive move that expands the funnel rather than harvests existing subscribers early. 5 That's a moat-widening action: more time spent in the app, more data, higher conversion probability downstream.

The DET exam business. The Duolingo English Test is a $59 exam that takes 60 minutes and can be taken at home. It undercuts TOEFL ($215+) and IELTS on both price and convenience, and its institutional acceptance list has grown to include universities in the US, UK, Canada, and Australia. This is a second-revenue stream tied to real-world credentialing — stickiness that extends beyond casual language practice.

At $111.36 (May 29, 2026 close), with market cap of $5.19B and EV of $4.03B: 2

The forward PE looks expensive in isolation. But the FCF-based multiples tell a different story: $370M of FCF in a business with $1.04B in revenue is a 35.6% FCF margin, and the market is pricing that stream at 12.5x on a TTM basis — a multiple that would be normal for a declining industrial company, not a software platform with 97% FCF CAGR and no debt.

The selloff repriced the stock from a speculative growth premium to a cash-flow-based value range in less than six months. Sell-side consensus is "Hold" with an average price target of $104.55 — below the current price — suggesting most analysts are still anchored on slowing growth rather than the cash yield that has emerged from the derating. 2

1. The engagement-first bet could delay monetization for longer than expected. Duolingo's 2026 strategy is deliberately prioritizing engagement growth over subscriber conversion. That's what caused the bookings guidance miss in Q1 (FY2026 bookings growth guided at ~10.5% vs. revenue growth of ~16.1%, implying declining booking-to-revenue conversion). If engagement improves but conversion lags into 2027, the business will still be fine, but free-cash-flow growth will decelerate before it reaccelerates.

2. DAU growth is decelerating. DAUs grew 21% in Q1 2026, down from 38% in Q1 2025. The app already has north of 50 million daily users — the law of large numbers is asserting itself. Maintaining double-digit DAU growth at this scale requires continuous product innovation to bring lapsed users back and to expand into new geographies.

3. Legal investigation risk. Multiple law firms have opened securities-fraud investigations into Duolingo following the Q1 earnings selloff, alleging that management guidance was misleading. 6 These investigations rarely result in material liability for well-governed companies, but they create headline risk and management distraction during what is already a transition year.

4. AI-native competition. Large language models now speak, correct grammar, and simulate conversation. ChatGPT, Claude, and purpose-built apps like Speak are competing for the same "want to practice a language with AI" use case. Duolingo's counter is that gamification, curriculum structure, and proven completion rates are what actually make learners stick — but this is a structural competitive threat that didn't exist three years ago.

5. Conversion rate sensitivity. The paid subscriber base of 12.5M sits atop 56.5M DAUs — a conversion rate of about 22%. If the free-tier expansion strategy pulls the conversion rate down before engagement and subsequent conversion can more than offset it, revenue per DAU will decline. Duolingo's FY2026 EBITDA guidance ($310M) implies management expects to hold margins, but execution risk is real.

Duolingo is a consumer software business that generated $370M in free cash flow on $1B in revenue, has no debt, holds $1B in net cash, and is priced at 12.5x TTM FCF — a multiple that reflects a market focused on near-term bookings deceleration rather than the cash economics of the underlying platform. The moat is real: 250 language courses, 56.5M daily users, gamification data that trains its AI, and a certified English exam business. The risk is also real: management is consciously trading near-term monetization for longer-term engagement, which compresses short-term earnings. For investors who care about what free cash flow does over three to five years, the setup is interesting. For investors who care about next quarter's bookings, it is not.